Passed surveys refer to those IUs that didn’t fail any impact or
surveillance surveys.
4 categories of “problem” IUs considered throughout (not mutually
exclusive):
- Failed impact surveys: failed at least 1 impact survey (TF >
5%)
- Persistent: failed at least 2 impact, never below 5% TF
- Failed surveillance surveys: failed at least 1 surveillance survey
(TF > 5%)
- Rebound: passed impact survey but failed surveillance survey
Summarise by category
For each IU, the median, 2.5% and 97.5% quantiles were calculated.
These quantiles were then averaged across all IUs in the respective
categories.
Passed surveys
| beta1996 |
0.0675241 |
0.0292626 |
0.1778853 |
| beta2000 |
0.0453407 |
0.0246494 |
0.0694646 |
| beta2010 |
0.0463910 |
0.0227371 |
0.0850335 |
| beta2020 |
0.0370703 |
0.0147294 |
0.0774441 |
| betareduction2000 |
-0.3882925 |
-0.6932053 |
-0.1041942 |
| betareduction2010 |
-0.3814495 |
-0.7077468 |
-0.1099961 |
| betareduction2020 |
-0.4592724 |
-0.7724160 |
-0.1567005 |
| eff_cov |
0.7390860 |
0.3895325 |
0.9541720 |
| mda_rounds |
0.0000000 |
0.0000000 |
5.0000000 |
Failed impact surveys
| beta1996 |
0.1368075 |
0.0836268 |
0.2645825 |
| beta2000 |
0.0849476 |
0.0581802 |
0.1406780 |
| beta2010 |
0.0901942 |
0.0621809 |
0.1343969 |
| beta2020 |
0.0716541 |
0.0386862 |
0.1126101 |
| betareduction2000 |
-0.3051410 |
-0.5997325 |
-0.0645869 |
| betareduction2010 |
-0.3642656 |
-0.6472770 |
-0.1226310 |
| betareduction2020 |
-0.4503289 |
-0.7449141 |
-0.1794181 |
| eff_cov |
0.7330953 |
0.3974022 |
0.9464806 |
| mda_rounds |
6.0000000 |
2.0000000 |
15.0000000 |
Persistent (failed at least 2 impact, never below 5% TF)
| beta1996 |
0.1573089 |
0.1024788 |
0.2906803 |
| beta2000 |
0.1213206 |
0.0827090 |
0.1945299 |
| beta2010 |
0.1142364 |
0.0856680 |
0.1653591 |
| beta2020 |
0.0850244 |
0.0479186 |
0.1286305 |
| betareduction2000 |
-0.4925589 |
-0.7277549 |
-0.2200454 |
| betareduction2010 |
-0.3228506 |
-0.5984778 |
-0.1025363 |
| betareduction2020 |
-0.4373248 |
-0.7304441 |
-0.1773022 |
| eff_cov |
0.6887384 |
0.3492101 |
0.9314825 |
| mda_rounds |
12.0000000 |
3.9500000 |
15.0000000 |
Failed surveillance surveys
| beta1996 |
0.1002894 |
0.0587708 |
0.1923575 |
| beta2000 |
0.0563758 |
0.0325857 |
0.0929904 |
| beta2010 |
0.0567877 |
0.0370085 |
0.0830548 |
| beta2020 |
0.0473338 |
0.0255905 |
0.0726202 |
| betareduction2000 |
-0.4453401 |
-0.7331980 |
-0.1486755 |
| betareduction2010 |
-0.4255939 |
-0.6892015 |
-0.1642519 |
| betareduction2020 |
-0.4927554 |
-0.7635229 |
-0.2167660 |
| eff_cov |
0.7118447 |
0.3911048 |
0.9333338 |
| mda_rounds |
5.0000000 |
2.0500000 |
11.9500000 |
Rebounds (passed impact, failed surveillance)
| beta1996 |
0.1034625 |
0.0580467 |
0.2077788 |
| beta2000 |
0.0542043 |
0.0299056 |
0.0834749 |
| beta2010 |
0.0508810 |
0.0274759 |
0.0801808 |
| beta2020 |
0.0441855 |
0.0216708 |
0.0718195 |
| betareduction2000 |
-0.4657185 |
-0.7473893 |
-0.1663904 |
| betareduction2010 |
-0.5046269 |
-0.7637502 |
-0.2162208 |
| betareduction2020 |
-0.5244174 |
-0.7961752 |
-0.2412629 |
| eff_cov |
0.7643509 |
0.4318633 |
0.9582316 |
| mda_rounds |
4.0000000 |
1.2000000 |
8.8000000 |
Comparison plots
- Plots of the intervals presented in the above tables
- Failed impact surveys (including persistent IUs) have higher median
beta on average. Not so much for failed surveillance surveys (including
rebound IUs)
- MDA rounds of rebounded IUs (passed impact but failed surveillance)
smaller than other “problem” categories

Summaries by baseline TF prevalence
Baseline TF vs baseline \(\beta\)
- Bimodality in the beta distribution for IUs with higher baseline
prevalence seems to driven by country
- Medians of effective coverage quite tight around the mean of the
prior distribution for IUs that passed all surveys, but not the case for
some of the “problem” categories, particularly at higher prevalence

Look at bimodality in passed IUs

Look at Ethiopia
## [1] "Colour by baseline prevalence"

## [1] "Colour by latest prevalence"

## [1] "Colour by batch ID"

Baseline TF vs \(\beta\) in
2020

Baseline TF vs overall \(\beta\)
reduction

Baseline TF vs effective coverage

Baseline TF vs number of MDA rounds

Coverage
- IUs that passed all surveys but had low coverage have high baseline
\(\beta\) and large reductions in beta
by 2020


Check trajectories for IUs that failed a surveillance survey or
rebounded
- Plotting IUs that failed a surveillance survey or rebounded to see
if their simulated trajectories are at the top end of the 0-5% TF
prevalence band
- For IUs where the rebounded data point wasn’t fit, the model still
hovers around 5% for most IUs. Although some IUs have some simulations
that approach 0%
- When the rebounded data point was fit, none of the IUs reach 0%
prevalence
Legend:
- Black points and grey ribbon: median and 95% interval
- Green horizontal line: 5%
- Black horizontal line: 0%
## [1] "rebound not fitted"

## [1] "rebound fitted"
